1 /* SPDX-License-Identifier: GPL-2.0 */ 1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* linux/arch/sparc/lib/memset.S: Sparc optimi !! 2 /* 3 * Copyright (C) 1991,1996 Free Software Found !! 3 * linux/arch/alpha/lib/memset.S 4 * Copyright (C) 1996,1997 Jakub Jelinek (jj@s << 5 * Copyright (C) 1996 David S. Miller (davem@c << 6 * 4 * 7 * Calls to memset returns initial %o0. Calls !! 5 * This is an efficient (and small) implementation of the C library "memset()" 8 * number of bytes not yet set if exception oc !! 6 * function for the alpha. 9 * clear_user. !! 7 * >> 8 * (C) Copyright 1996 Linus Torvalds >> 9 * >> 10 * This routine is "moral-ware": you are free to use it any way you wish, and >> 11 * the only obligation I put on you is a moral one: if you make any improvements >> 12 * to the routine, please send me your improvements for me to use similarly. >> 13 * >> 14 * The scheduling comments are according to the EV5 documentation (and done by >> 15 * hand, so they might well be incorrect, please do tell me about it..) 10 */ 16 */ 11 << 12 #include <linux/export.h> 17 #include <linux/export.h> 13 #include <asm/ptrace.h> !! 18 .set noat >> 19 .set noreorder >> 20 .text >> 21 .globl memset >> 22 .globl __memset >> 23 .globl ___memset >> 24 .globl __memset16 >> 25 .globl __constant_c_memset >> 26 >> 27 .ent ___memset >> 28 .align 5 >> 29 ___memset: >> 30 .frame $30,0,$26,0 >> 31 .prologue 0 >> 32 >> 33 and $17,255,$1 /* E1 */ >> 34 insbl $17,1,$17 /* .. E0 */ >> 35 bis $17,$1,$17 /* E0 (p-c latency, next cycle) */ >> 36 sll $17,16,$1 /* E1 (p-c latency, next cycle) */ >> 37 >> 38 bis $17,$1,$17 /* E0 (p-c latency, next cycle) */ >> 39 sll $17,32,$1 /* E1 (p-c latency, next cycle) */ >> 40 bis $17,$1,$17 /* E0 (p-c latency, next cycle) */ >> 41 ldq_u $31,0($30) /* .. E1 */ >> 42 >> 43 .align 5 >> 44 __constant_c_memset: >> 45 addq $18,$16,$6 /* E0 */ >> 46 bis $16,$16,$0 /* .. E1 */ >> 47 xor $16,$6,$1 /* E0 */ >> 48 ble $18,end /* .. E1 */ >> 49 >> 50 bic $1,7,$1 /* E0 */ >> 51 beq $1,within_one_quad /* .. E1 (note EV5 zero-latency forwarding) */ >> 52 and $16,7,$3 /* E0 */ >> 53 beq $3,aligned /* .. E1 (note EV5 zero-latency forwarding) */ >> 54 >> 55 ldq_u $4,0($16) /* E0 */ >> 56 bis $16,$16,$5 /* .. E1 */ >> 57 insql $17,$16,$2 /* E0 */ >> 58 subq $3,8,$3 /* .. E1 */ >> 59 >> 60 addq $18,$3,$18 /* E0 $18 is new count ($3 is negative) */ >> 61 mskql $4,$16,$4 /* .. E1 (and possible load stall) */ >> 62 subq $16,$3,$16 /* E0 $16 is new aligned destination */ >> 63 bis $2,$4,$1 /* .. E1 */ >> 64 >> 65 bis $31,$31,$31 /* E0 */ >> 66 ldq_u $31,0($30) /* .. E1 */ >> 67 stq_u $1,0($5) /* E0 */ >> 68 bis $31,$31,$31 /* .. E1 */ >> 69 >> 70 .align 4 >> 71 aligned: >> 72 sra $18,3,$3 /* E0 */ >> 73 and $18,7,$18 /* .. E1 */ >> 74 bis $16,$16,$5 /* E0 */ >> 75 beq $3,no_quad /* .. E1 */ >> 76 >> 77 .align 3 >> 78 loop: >> 79 stq $17,0($5) /* E0 */ >> 80 subq $3,1,$3 /* .. E1 */ >> 81 addq $5,8,$5 /* E0 */ >> 82 bne $3,loop /* .. E1 */ >> 83 >> 84 no_quad: >> 85 bis $31,$31,$31 /* E0 */ >> 86 beq $18,end /* .. E1 */ >> 87 ldq $7,0($5) /* E0 */ >> 88 mskqh $7,$6,$2 /* .. E1 (and load stall) */ >> 89 >> 90 insqh $17,$6,$4 /* E0 */ >> 91 bis $2,$4,$1 /* .. E1 */ >> 92 stq $1,0($5) /* E0 */ >> 93 ret $31,($26),1 /* .. E1 */ >> 94 >> 95 .align 3 >> 96 within_one_quad: >> 97 ldq_u $1,0($16) /* E0 */ >> 98 insql $17,$16,$2 /* E1 */ >> 99 mskql $1,$16,$4 /* E0 (after load stall) */ >> 100 bis $2,$4,$2 /* E0 */ >> 101 >> 102 mskql $2,$6,$4 /* E0 */ >> 103 mskqh $1,$6,$2 /* .. E1 */ >> 104 bis $2,$4,$1 /* E0 */ >> 105 stq_u $1,0($16) /* E0 */ >> 106 >> 107 end: >> 108 ret $31,($26),1 /* E1 */ >> 109 .end ___memset >> 110 EXPORT_SYMBOL(___memset) >> 111 EXPORT_SYMBOL(__constant_c_memset) >> 112 >> 113 .align 5 >> 114 .ent __memset16 >> 115 __memset16: >> 116 .prologue 0 >> 117 >> 118 inswl $17,0,$1 /* E0 */ >> 119 inswl $17,2,$2 /* E0 */ >> 120 inswl $17,4,$3 /* E0 */ >> 121 or $1,$2,$1 /* .. E1 */ >> 122 inswl $17,6,$4 /* E0 */ >> 123 or $1,$3,$1 /* .. E1 */ >> 124 or $1,$4,$17 /* E0 */ >> 125 br __constant_c_memset /* .. E1 */ 14 126 15 /* Work around cpp -rob */ !! 127 .end __memset16 16 #define ALLOC #alloc !! 128 EXPORT_SYMBOL(__memset16) 17 #define EXECINSTR #execinstr << 18 #define EX(x,y,a,b) << 19 98: x,y; << 20 .section .fixup,ALLOC,EXECINSTR; << 21 .align 4; << 22 99: retl; << 23 a, b, %o0; << 24 .section __ex_table,ALLOC; << 25 .align 4; << 26 .word 98b, 99b; << 27 .text; << 28 .align 4 << 29 << 30 #define STORE(source, base, offset, n) << 31 98: std source, [base + offset + n]; << 32 .section .fixup,ALLOC,EXECINSTR; << 33 .align 4; << 34 99: ba 30f; << 35 sub %o3, n - offset, %o3; << 36 .section __ex_table,ALLOC; << 37 .align 4; << 38 .word 98b, 99b; << 39 .text; << 40 .align 4; << 41 << 42 #define STORE_LAST(source, base, offset, n) << 43 EX(std source, [base - offset - n], << 44 add %o1, offset + n); << 45 << 46 /* Please don't change these macros, unless yo << 47 * in the .fixup section below as well. << 48 * Store 64 bytes at (BASE + OFFSET) using val << 49 #define ZERO_BIG_BLOCK(base, offset, source) << 50 STORE(source, base, offset, 0x00); << 51 STORE(source, base, offset, 0x08); << 52 STORE(source, base, offset, 0x10); << 53 STORE(source, base, offset, 0x18); << 54 STORE(source, base, offset, 0x20); << 55 STORE(source, base, offset, 0x28); << 56 STORE(source, base, offset, 0x30); << 57 STORE(source, base, offset, 0x38); << 58 << 59 #define ZERO_LAST_BLOCKS(base, offset, source) << 60 STORE_LAST(source, base, offset, 0x38) << 61 STORE_LAST(source, base, offset, 0x30) << 62 STORE_LAST(source, base, offset, 0x28) << 63 STORE_LAST(source, base, offset, 0x20) << 64 STORE_LAST(source, base, offset, 0x18) << 65 STORE_LAST(source, base, offset, 0x10) << 66 STORE_LAST(source, base, offset, 0x08) << 67 STORE_LAST(source, base, offset, 0x00) << 68 << 69 .text << 70 .align 4 << 71 << 72 .globl __bzero_begin << 73 __bzero_begin: << 74 << 75 .globl __bzero << 76 .type __bzero,#function << 77 .globl memset << 78 EXPORT_SYMBOL(__bzero) << 79 EXPORT_SYMBOL(memset) << 80 memset: << 81 mov %o0, %g1 << 82 mov 1, %g4 << 83 and %o1, 0xff, %g3 << 84 sll %g3, 8, %g2 << 85 or %g3, %g2, %g3 << 86 sll %g3, 16, %g2 << 87 or %g3, %g2, %g3 << 88 b 1f << 89 mov %o2, %o1 << 90 3: << 91 cmp %o2, 3 << 92 be 2f << 93 EX(stb %g3, [%o0], sub %o1, 0) << 94 << 95 cmp %o2, 2 << 96 be 2f << 97 EX(stb %g3, [%o0 + 0x01], sub %o1, 1) << 98 << 99 EX(stb %g3, [%o0 + 0x02], sub %o1, 2) << 100 2: << 101 sub %o2, 4, %o2 << 102 add %o1, %o2, %o1 << 103 b 4f << 104 sub %o0, %o2, %o0 << 105 << 106 __bzero: << 107 clr %g4 << 108 mov %g0, %g3 << 109 1: << 110 cmp %o1, 7 << 111 bleu 7f << 112 andcc %o0, 3, %o2 << 113 << 114 bne 3b << 115 4: << 116 andcc %o0, 4, %g0 << 117 << 118 be 2f << 119 mov %g3, %g2 << 120 << 121 EX(st %g3, [%o0], sub %o1, 0) << 122 sub %o1, 4, %o1 << 123 add %o0, 4, %o0 << 124 2: << 125 andcc %o1, 0xffffff80, %o3 ! Now << 126 be 9f << 127 andcc %o1, 0x78, %o2 << 128 10: << 129 ZERO_BIG_BLOCK(%o0, 0x00, %g2) << 130 subcc %o3, 128, %o3 << 131 ZERO_BIG_BLOCK(%o0, 0x40, %g2) << 132 bne 10b << 133 add %o0, 128, %o0 << 134 << 135 orcc %o2, %g0, %g0 << 136 9: << 137 be 13f << 138 andcc %o1, 7, %o1 << 139 << 140 srl %o2, 1, %o3 << 141 set 13f, %o4 << 142 sub %o4, %o3, %o4 << 143 jmp %o4 << 144 add %o0, %o2, %o0 << 145 << 146 ZERO_LAST_BLOCKS(%o0, 0x48, %g2) << 147 ZERO_LAST_BLOCKS(%o0, 0x08, %g2) << 148 13: << 149 be 8f << 150 andcc %o1, 4, %g0 << 151 << 152 be 1f << 153 andcc %o1, 2, %g0 << 154 << 155 EX(st %g3, [%o0], and %o1, 7) << 156 add %o0, 4, %o0 << 157 1: << 158 be 1f << 159 andcc %o1, 1, %g0 << 160 << 161 EX(sth %g3, [%o0], and %o1, 3) << 162 add %o0, 2, %o0 << 163 1: << 164 bne,a 8f << 165 EX(stb %g3, [%o0], and %o1, 1) << 166 8: << 167 b 0f << 168 nop << 169 7: << 170 be 13b << 171 orcc %o1, 0, %g0 << 172 << 173 be 0f << 174 8: << 175 add %o0, 1, %o0 << 176 subcc %o1, 1, %o1 << 177 bne 8b << 178 EX(stb %g3, [%o0 - 1], add %o1, 1) << 179 0: << 180 andcc %g4, 1, %g0 << 181 be 5f << 182 nop << 183 retl << 184 mov %g1, %o0 << 185 5: << 186 retl << 187 clr %o0 << 188 << 189 .section .fixup,#alloc,#execinstr << 190 .align 4 << 191 30: << 192 and %o1, 0x7f, %o1 << 193 retl << 194 add %o3, %o1, %o0 << 195 129 196 .globl __bzero_end !! 130 memset = ___memset 197 __bzero_end: !! 131 __memset = ___memset >> 132 EXPORT_SYMBOL(memset) >> 133 EXPORT_SYMBOL(__memset)
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.